When the AI Tutor Sits Unused: Why Access Is Not the Same as Engagement

Summary for Educators

Based on Matt Barnum

The problem with an AI tutor: Students don’t want to use it

Chalkbeat | August 25, 2026
Chalkbeat Ideas — reporting on new research examining student use of Khan Academy’s AI-powered tutor, Khanmigo. (Chalkbeat⁠)

🔵 THE BIG IDEA

Schools have spent the past several years hearing an enticing promise: Artificial intelligence could provide every student with something resembling a personal tutor—available anytime, infinitely patient, and able to customize explanations to individual needs.

But a fundamental problem is emerging.

Many students simply don’t use the tutor.

A new two-year study involving middle school students using Khan Academy found that although access to its AI-powered Khanmigo tutor was nearly universal, meaningful engagement was surprisingly limited. Students used Khanmigo on only about one-third of the days they were working in Khan Academy. When they did engage, some sent off-topic messages or attempted to persuade the chatbot to simply provide answers. (Chalkbeat⁠)

The lesson for school leaders is important: Providing students with powerful technology does not automatically produce powerful learning.

AI may be capable of tutoring. The more difficult question is whether students are motivated to be tutored.

🔵 KEY TAKEAWAYS FOR EDUCATORS

• Access does not equal engagement. Putting an AI tutor on every student’s screen is very different from getting students to use it thoughtfully. Researchers summarized the problem succinctly: access was nearly universal, but engagement remained thin. (Chalkbeat⁠)
• Students often choose the path of least resistance. A good tutor asks questions, provides hints, and requires students to think. But students who simply want to finish an assignment may prefer a tool that gives them an answer immediately.
• Motivation remains fundamentally human. AI can explain a quadratic equation or provide reading support, but it cannot automatically make a reluctant student want to persist through difficult work.
• Human support still matters. Earlier Stanford research found that providing human encouragement increased students’ AI-tutor use only modestly—by roughly one to four additional minutes per week. Many students never logged in at all. (National Student Support Accelerator⁠)
• AI works better when embedded in learning. Khan Academy has subsequently redesigned Khanmigo so that AI assistance is woven more directly into the learning experience rather than waiting passively for students to ask for help. (Chalkbeat⁠)
• Productive struggle must remain part of the design. When technology makes avoiding effort easier than learning, convenience can work against achievement.

◻️ WHY IT MATTERS

School leaders evaluating AI products should be careful about asking only, “What can this technology do?”

An equally important question is:

“What will students actually do with it?”

Sal Khan himself has acknowledged that Khanmigo was initially a “non-event” for many students because they simply did not use it very much. His analogy is revealing: Place an excellent tutor in the back of a classroom and wait for students to voluntarily approach for help. Some will. Many will not. (Chalkbeat⁠)

That distinction matters enormously for districts considering expensive AI initiatives.

A sophisticated tutoring system with low student engagement may produce less learning than a simpler instructional intervention that students and teachers actually use consistently.

Technology adoption therefore needs to be viewed as an instructional-design challenge, not merely a purchasing decision.

🟢 LEADERSHIP ACTION STEPS

✓ Ask students first. Before purchasing an AI tutoring platform, conduct student focus groups and observe how students actually seek help when they become confused.

✓ Pilot before scaling. Test AI tutoring with a small group of teachers and students and examine actual usage—not simply login numbers.

✓ Measure learning behavior. Determine whether students are asking questions, revising thinking, practicing difficult concepts, and persisting—not merely opening the application.

✓ Integrate AI into instruction. Avoid treating an AI tutor as an optional website sitting beside the curriculum. Connect its use to classroom lessons, teacher feedback, and specific learning goals.

✓ Protect productive struggle. Select tools that coach students toward answers rather than eliminating the thinking required to reach them.

✓ Keep humans in the loop. Teachers, tutors, classmates, and mentors continue to provide something technology struggles to reproduce: encouragement, accountability, relationships, and the ability to recognize when a student has stopped trying.

🟡 LEADER REFLECTION

If we gave every student in our school an excellent AI tutor tomorrow, how many would actually use it to learn?

And perhaps the more important question:

Are we designing AI around what technology can do—or around how students actually learn?

The emerging research offers school leaders a useful warning. AI tutoring may indeed become remarkably capable. But education has never been simply an information problem.

Sometimes the most important ingredient in tutoring isn’t the tutor.

It’s getting the student to want to learn.

Original article

The problem with an AI tutor: Students don’t want to use it

Chalkbeat | August 25, 2026

------------------------------

Prepared with the assistance of AI software

OpenAI. (2026). ChatGPT (5.2) [Large language model]. https://chat.openai.com